In particle filter simultaneous localization and mapping method (SLAM) for mobile robots, linearization of nonlinear systems and derivation of the Jacobian matrices lead to large computing and system state estimation problem of low accuracy. To solve this problem，a SLAM method based on interval analysis and unscented particle filter was proposed. The number of particles was seriously reduced based on the interval analysis for pose estimation. The nature landmarks were updated through unscented Kalman filter to improve map estimation accuracy. This approach improved the state estimation accuracy and computational effort with a smaller number of particles than previous approaches. The results indicated that the proposed method was valid.